several errors first report

[Soldorado]

My language isn't English. I hope that makes sense. The chat link has expired, and I can't find a way to report errors. I'm using nextstep-1 on stepfun.ai.

(1)

The mask commands to try to prevent a character from changing position in the final images are buggy. Sometimes white rectangles appear, resembling sheets of paper, somewhere within the mask.

(2)

The commands to prevent unwanted 2D artistic styles from appearing in the final result don't work at all. I find it impossible to prevent drawings in various styles from appearing. When I try to only display still images, this is the latest one I've tried, and it's still in an experimental state.

--!no_2d_output 

(3)

I cannot fully test the control_mode, which works similarly to controlnet, as it cannot prevent drawings from appearing in the final result.

(4)

There are certain descriptions that redirect you to generate drawings in the prompt. For example, those about fantasy, medieval times with ethnic mythology, etc., even if you specify that a photograph be provided.

(5)

The drawing-to-photo conversation is the one that needs the most work. In some tests with Control Mode, it worked well for me. But if you have a drawing of a bear as an example and you want to convert it to a photo with certain keywords, it completely changes the design and even its position. Sometimes a prompt like (bear-style fur) works, but in most cases it doesn't.

(6)

The commands for selecting the ethnic type of the person to convert to a photograph. Only the one for specifying ethnicity works. The ones for changing eyes, skin color, and so on. They always send you drawings in the final result.

(7)

Conversations from drawings of people or animals to real full-body photographs usually work well, but the limit is first-person faces, that is, those very close to the character, such as a portrait or profile drawing. It always generates drawings at the end, even if you specify otherwise.

(8)

When a person in a drawing is of medium height and you want to convert it to a photograph, you inevitably end up with children instead of adults in most cases.

(9)

When you want to convert drawings like fantasy dwarves from Lord of the Rings into real photography, depending on the pose in the drawing and the design features. The resulting artwork is created.

I've only had one successful result, a dwarf photograph. I've also had success with a drawing of a goblin.

(10)

The model needs an algorithm to convert, for example, the faces into drawings and convert them to real photographs with the same features. The goblin's results were the same, with the same positions, since I used the drawing fidelity command adjusted by numbers. And the goblin's skin came out almost as realistic as the photograph.

(11)

I use commands to avoid the bug with smears on magenta and cyan textures, but sometimes the textures of those colors still get smeared.